feat: Ctrl+l 快速切換 5 碼和 6 碼輸入時的拆分

例如，想要用 dafju 輸入「打句」，首選切分卻是「打飯」（da'fju），則可
以按 Ctrl+l 快速切換爲 daf'ju。再按一次 Ctrl+l 可以回到原來的切分。

輸入爲 6 碼時，可在 2-2-2 和 3-3 兩種切分中切換。

lua/moran_processor.lua

@@ -0,0 +1,143 @@
+-- moran_processor.lua
+-- Synopsis: 適用於魔然方案默認模式的按鍵處理器
+-- Author: ksqsf
+-- License: MIT license
+-- Version: 0.2
+
+-- 主要功能：
+-- 1. 選擇第二個首選項，但可用於跳過 emoji 濾鏡產生的候選
+-- 2. 快速切換強制切分
+
+-- ChangeLog:
+--  0.2.0: 增加快速切換切分的能力，因而從 moran_semicolon_processor 更名爲 moran_processor
+--  0.1.5: 修復獲取 candidate_count 的邏輯
+--  0.1.4: 數字也增加到條件裏
+
+local moran = require("moran")
+
+local kReject = 0
+local kAccepted = 1
+local kNoop = 2
+
+local function semicolon_processor(key_event, env)
+   local context = env.engine.context
+
+   if key_event.keycode ~= 0x3B then
+      return kNoop
+   end
+
+   local composition = context.composition
+   if composition:empty() then
+      return kNoop
+   end
+
+   local segment = composition:back()
+   local menu = segment.menu
+   local page_size = env.engine.schema.page_size
+
+   -- Special cases: for 'ovy' and 快符, just send ';'
+   if context.input:find('^ovy') or context.input:find('^;') then
+      return kNoop
+   end
+
+   -- Special case: if there is only one candidate, just select it!
+   local candidate_count = menu:prepare(page_size)
+   if candidate_count == 1 then
+      context:select(0)
+      return kAccepted
+   end
+
+   -- If it is not the first page, simply send 2.
+   local selected_index = segment.selected_index
+   if selected_index >= page_size then
+      local page_num = selected_index // page_size
+      context:select(page_num * page_size + 1)
+      return kAccepted
+   end
+
+   -- First page: do something more sophisticated.
+   local i = 1
+   while i < page_size do
+      local cand = menu:get_candidate_at(i)
+      if cand == nil then
+         context:select(1)
+         return kNoop
+      end
+      local cand_text = cand.text
+      local codepoint = utf8.codepoint(cand_text, 1)
+      if moran.unicode_code_point_is_chinese(codepoint) -- 漢字
+         or (codepoint >= 97 and codepoint <= 122)      -- a-z
+         or (codepoint >= 65 and codepoint <= 90)       -- A-Z
+         or (codepoint >= 48 and codepoint <= 57 and cand.type ~= "simplified") -- 0-9
+      then
+         context:select(i)
+         return kAccepted
+      end
+      i = i + 1
+   end
+
+   -- No good candidates found. Just select the second candidate.
+   context:select(1)
+   return kAccepted
+end
+
+local function force_segmentation_processor(key_event, env)
+   if not (key_event:ctrl() and key_event.keycode == 0x6c) then  -- ctrl+l
+      return kNoop
+   end
+
+   local composition = env.engine.context.composition
+   if composition:empty() then
+      return kNoop
+   end
+
+   local seg = composition:back()
+   local cand = seg:get_selected_candidate()
+   if cand == nil then
+      return kNoop
+   end
+
+   local ctx = env.engine.context
+   local input = ctx.input:sub(seg._start + 1, seg._end)
+   local preedit = cand.preedit
+
+   local raw = input:gsub("'", "")  -- 不帶 ' 分隔符的輸入
+
+   if preedit:match("^[a-z][a-z][ '][a-z][a-z][a-z]$") or input:match("^[a-z][a-z]'[a-z][a-z][a-z]$") then  -- 2-3
+      ctx.input = ctx.input:sub(1, seg._start) .. raw:sub(1,3) .. "'" .. raw:sub(4,5) .. ctx.input:sub(seg._end + 1, -1)
+   elseif preedit:match("^[a-z][a-z][a-z][ '][a-z][a-z]$") or input:match("^[a-z][a-z][a-z]'[a-z][a-z]$") then  -- 3-2
+      ctx.input = ctx.input:sub(1, seg._start) .. raw:sub(1,2) .. "'" .. raw:sub(3,5) .. ctx.input:sub(seg._end + 1, -1)
+   elseif preedit:match("^[a-z][a-z][a-z][ '][a-z][a-z][a-z]$") or input:match("^[a-z][a-z][a-z]'[a-z][a-z][a-z]$") then -- 3-3
+      ctx.input = ctx.input:sub(1, seg._start) .. raw:sub(1,2) .. "'" .. raw:sub(3,4) .. "'" .. raw:sub(5,6) .. ctx.input:sub(seg._end + 1, -1)
+   elseif preedit:match("^[a-z][a-z][ '][a-z][a-z][ '][a-z][a-z]$") or input:match("^[a-z][a-z]'[a-z][a-z]'[a-z][a-z]$") then  -- 2-2-2
+      ctx.input = ctx.input:sub(1, seg._start) .. raw:sub(1,3) .. "'" .. raw:sub(4,6) .. ctx.input:sub(seg._end + 1, -1)
+   end
+
+   return kAccepted
+end
+
+return {
+   init = function(env)
+      env.processors = {
+         semicolon_processor,
+         force_segmentation_processor,
+      }
+   end,
+
+   fini = function(env)
+   end,
+
+   func = function(key_event, env)
+      if key_event:release() then
+         return kNoop
+      end
+
+      for _, processor in pairs(env.processors) do
+         local res = processor(key_event, env)
+         if res == kAccepted or res == kRejected then
+            return res
+         end
+      end
+      return kNoop
+   end
+}

moran.schema.yaml

@@ -52,7 +52,7 @@ engine:
   processors:
     - key_binder
     - lua_processor@*moran_pin*pin_processor
-    - lua_processor@*moran_semicolon_processor
+    - lua_processor@*moran_processor
     - ascii_composer
     - recognizer
     - speller
@@ -65,8 +65,6 @@ engine:
     - matcher
     - matcher@recognizer_secondary
     - affix_segmentor@japanese_o
-    - affix_segmentor@force_trad
-    - affix_segmentor@force_simp
     - abc_segmentor
     - punct_segmentor
     - fallback_segmentor
@@ -299,14 +297,6 @@ reverse_lookup:
     - reverse_zrlf
     - reverse_bopomofo
 
-force_simp:
-  suffix: '/j'
-  tag: force_simp
-
-force_trad:
-  suffix: '/f'
-  tag: force_trad
-
 punctuator:
   import_preset: symbols
 
@@ -420,3 +410,4 @@ __include: moran:/octagram/enable_for_sentence
 # 若要禁用，可直接刪掉上面的 __include，或在 custom 文件中寫入：
 # patch:
 #   __include: moran:/octagram/disable
+

tests/moran.test.yaml

@@ -90,3 +90,18 @@ deploy:
         assert: eq(cand[1].text, '連接') and eq(cand[1].comment, "📌")
       - send: 'lmjx'
         assert: eq(cand[1].text, '連接') and eq(cand[1].comment, "📌")
+
+  force_segmentation:
+    tests:
+      - send: 'dafju'
+        assert: eq(preedit, "da fju")
+      - send: 'dafju{Control+l}'
+        assert: eq(preedit, "daf'ju")
+      - send: 'dafju{Control+l}{Control+l}'
+        assert: eq(preedit, "da'fju")
+      - send: 'dajwhk'
+        assert: eq(preedit, 'da jw hk')
+      - send: 'dajwhk{Control+l}'
+        assert: eq(preedit, "daj'whk")
+      - send: 'dajwhk{Control+l}{Control+l}'
+        assert: eq(preedit, "da'jw'hk")